metatron: Please provide a URL to reproduce / see the problem.
I assume this is some page somewhere on http://dumps.wikimedia.org/

metatron wrote:

https://dumps.wikimedia.org/other/pagecounts-raw/2014/2014-05/

The files are still there on gadolinium, so it looks like we're not loosing data,
but "only" copying the data files around is failing.

Change 136735 had a related patch set uploaded by QChris:
Provide default shell to datasets user

https://gerrit.wikimedia.org/r/136735

Change 136735 abandoned by QChris:
Provide default shell to datasets user

Reason:
Beaten by Change 136734

https://gerrit.wikimedia.org/r/136735

It seems Ops' cleanup around accounts reset the datasets user's shell,
so rsync failed. Apergos set rsync's shell again in

https://gerrit.wikimedia.org/r/#/c/136734

. That should fix the issue upon the next cron run and bring the missing
files over.
I'll report back when that happened.

The fix worked.
Files are back.
No missing data.

otto wrote:

Thanks Christian and Ariel (and all)!